摘要
鱼类的运动行为能反映出与水流的相适应程度,从而为鱼道设计提供依据,鱼类目标检测是研究鱼类运动行为的前提,因此如何高效准确的在复杂场景中检测和定位鱼类成为关键。文章主要研究了基于深度学习的目标检测YOLOv3算法,实现了在鱼道复杂场景中鱼的实时检测和计数,保证了检测的速度的同时也提高了准确率。实验表明,该方法能够高效的准确检测出鱼道中鱼类,实时记录鱼道中鱼的数目以及鱼的长度信息,可为后续研究鱼类行为提供数据基础,有助于鱼道设计。
The fish’s movement behavior can reflect the degree of adaptability to water flow, thus providing a basis for designing fishway. Fish target detection is a prerequisite for studying fish movement behavior, so how to efficiently and accurately detect fish in complex scenes becomes the key. This paper mainly studies the target detection YOLOv3 algorithm based on deep learning, realizing the real-time detection of fish and counting the number of fish in the complex scene of fishway, ensuring the speed of detection and improving the accuracy. Experiments show that the method can accurately and accurately detect fish in the fishway, record the number of fish in the fish pass and the length of the fish in real time,which can provide a data basis for the subsequent study of fish behavior and contribute to fishway design.
出处
《信息通信》
2019年第2期67-69,共3页
Information & Communications
关键词
鱼类
鱼道
检测
实时
Fish
Fishway
Detection
Real-time
